1

Topic: Service Discovery

Greetings, Are less than ten different projects which with each other communicate. One server on # One client application on # for workstations Some different servers on Java the Main protocol of dialogue - SOAP. But it is sometimes simple XML on HTTP. It would be desirable to simplify a configuration and  various pairs a host \port anywhere. Advise something from practice. Here that was possible  in Google - Eureka, ZooKeeper, Consul and similar - are able to do Discovery, but all remaining features are oriented on a cluster, plus it is necessary to contain a certain central register. - WS-Discovery, like as, a standard piece, but for a long time all forgotten and very few people generally . - UPnP, like as, approaches, with that only a difference that the protocol is guided by devices in a network, than on services more. It would be desirable something idle time and . Advise, please.

2

Re: Service Discovery

Hello, Blazkowicz, you wrote: B> Greetings, B> Are less than ten different projects which with each other communicate. B> one server on # B> One client application on # for workstations B> Some different servers on Java B> the Main protocol of dialogue - SOAP. But it is sometimes simple XML on HTTP. B> It would be desirable to simplify a configuration and  various pairs a host \port anywhere. B> advise something from practice. Here that was possible  in Google B> - Eureka, ZooKeeper, Consul and similar - are able to do Discovery, but all remaining features are oriented on a cluster, plus it is necessary to contain a certain central register. B> - WS-Discovery, like as, a standard piece, but for a long time all forgotten and very few people generally . B> - UPnP, like as, approaches, with that only a difference that the protocol is guided by devices in a network, than on services more. B> It would be desirable something idle time and . Advise, please. In practice and did not reach (while), but like how for such tasks is Enterprise service bus

3

Re: Service Discovery

Hello, Blazkowicz, you wrote: B> - UPnP, like as, approaches, with that only a difference that the protocol is guided by devices in a network, than on services more. UPnP it just is guided by services, however overkill, most likely, therefore as your services are hardly inscribed in standard circuits UPnP. It is possible to take a piece therefrom, namely SSDP (Simple Service Discovery Protocol). Here is  for.NET https://github.com/Yortw/RSSDP, is for a C https://github.com/zlargon/lssdp For Java libraries I do not know, but  simple as valenoks, there is a heap of implementations as an example.

4

Re: Service Discovery

Hello, Blazkowicz, you wrote: B> - UPnP, like as, approaches, with that only a difference that the protocol is guided by devices in a network, than on services more. UPnP - an excellent variant. Devices in a network it is not simple so, they to give services. Actually, to you it is necessary SSDP It is possible still ZeroConf to look at everyones, but I, am better it or is worse. I was picked with SSDP, all devices at me in a house network respond - a router (old and new), NAS, computers. I do not know, whether there is at all this economy support ZeroConf and similar. Still plus - the Windows such devices itself see and display in "the Network" root, it is possible any  to look without auxiliary tools, to pass to the device web page

5

Re: Service Discovery

Hello, andrey.desman, you wrote: B>> - UPnP, like as, approaches, with that only a difference that the protocol is guided by devices in a network, than on services more. AD> UPnP it just is guided by services, however overkill, most likely, therefore as your services are hardly inscribed in standard circuits UPnP. Behind prescription I can be mistaken, but UPnP, besides several protocols, describes any dial-up of standard services/devices. But nobody hinders to create the device service, is not mandatory it somewhere to register. UUID generated own, and use, hardly with something it will be intersected

6

Re: Service Discovery

Hello, Marty, you wrote: M> Behind prescription I can be mistaken, but UPnP, besides several protocols, describes any dial-up of standard services/devices. But nobody hinders to create the device service, is not mandatory it somewhere to register. UUID generated own, and use, hardly with something it will be intersected For this reason upnp it is not necessary, and its part - SSDP is necessary.

7

Re: Service Discovery

Hello, andrey.desman, you wrote: M>> Behind prescription I can be mistaken, but UPnP, besides several protocols, describes any dial-up of standard services/devices. But nobody hinders to create the device service, is not mandatory it somewhere to register. UUID generated own, and use, hardly with something it will be intersected AD> For this reason upnp it is not necessary, and its part - SSDP is necessary. And I another wrote something?

8

Re: Service Discovery

Hello, Blazkowicz, you wrote: B> Greetings, B> Are less than ten different projects which with each other communicate. B> one server on # B> One client application on # for workstations B> Some different servers on Java B> the Main protocol of dialogue - SOAP. But it is sometimes simple XML on HTTP. B> It would be desirable to simplify a configuration and  various pairs a host \port anywhere. B> advise something from practice. Here that was possible  in Google B> - Eureka, ZooKeeper, Consul and similar - are able to do Discovery, but all remaining features are oriented on a cluster, plus it is necessary to contain a certain central register. B> - WS-Discovery, like as, a standard piece, but for a long time all forgotten and very few people generally . B> - UPnP, like as, approaches, with that only a difference that the protocol is guided by devices in a network, than on services more. B> It would be desirable something idle time and . Advise, please. Consul. The centralized register is not necessary - use services with  on . I.e. the circuit approximately the such: - to deliver a cluster of the consul, it is possible from one  - on each server to deliver the client of the consul, many resources are not required to it - each application at start is registered in the consul of times in 10 seconds sends in localhost:8500 (the local consul) healthcheck with TTL 15 seconds. If application to stop, its record in the consul "becomes rotten" in 15 seconds. Then in the consul always there will be an actual information on live applications, their hosts and ports.  about services it is requested too through localhost:8500 so in application even it is not necessary to sew up the address of the consul. edit: and here for the client application the server address is better to distribute through DNS, well or   nginx c autoupdate of a config from the consul.

9

Re: Service Discovery

Hello, Marty, you wrote: AD>> For this reason upnp it is not necessary, and its part - SSDP is necessary. M> And I another wrote something? Yes, you wrote about upnp.

10

Re: Service Discovery

Hello, Blazkowicz, you wrote: B> It would be desirable something idle time and . Advise, please. If without  it would be possible, all interaction between services to do through the broker of messages. That is on a subscription to messages or their sending. But as at you , here a question not architecture, and developments. It is possible to tear all on a cloud in docker containers. And a location where what to send to implement through API  these containers. For example I now through kubernates try to make a similar problem. All development I do on kubernates a cluster, uniform for only. On this platform I have other clusters united for example through hazelcast. Well and hazelcast  thus that he IP  which it will unite, learned through Kubernates API. Almost completed already.

11

Re: Service Discovery

Hello, elmal, you wrote: E> But as at you , here a question not architecture, and developments. It is possible to tear all on a cloud in docker containers. It seems that I did not specify the most important thing. At me a shrink-wrapped software product. Therefore and it is so much fuss with development. A cloud, docker, here especially do not help.

12

Re: Service Discovery

Hello, sr_dev, you wrote: _> In practice and did not reach (while), but like how for such tasks is Enterprise service bus Is not present. ESB it is a transport communication center. Yes, he partially solves also my problem. But at us transport is already implemented.

13

Re: Service Discovery

Hello, scf, you wrote: scf> - each application at start is registered in the consul of times in 10 seconds sends in localhost:8500 (the local consul) healthcheck with TTL 15 seconds. If application to stop, its record in the consul "becomes rotten" in 15 seconds. It is all it is already implemented in the client code? scf> edit: and here for the client application the server address is better to distribute through DNS, well or   nginx c autoupdate of a config from the consul. Why? Need to twist DNS at clients only complicates . At us a shrink-wrapped software product. I am sorry, if at once did not specify.

14

Re: Service Discovery

Hello, andrey.desman, you wrote: AD> It is possible to take a piece therefrom, namely SSDP (Simple Service Discovery Protocol). Yes, after  I at it am declined. But wanted to consult. Like as, microservices  . Someone can understood.

15

Re: Service Discovery

Hello, Blazkowicz, you wrote: B> It seems that I did not specify the most important thing. At me a shrink-wrapped software product. Therefore and it is so much fuss with development. A cloud, docker, here especially do not help. At me too the box. Thus a cloud . Accordingly it is necessary on necessary  to deliver corresponding  and the docker to adjust it as it is necessary. Administrators for a week consulted With me. And further it is unrolled on time two and it is fine . As I will complete at last, there will be the full hair dryer .

16

Re: Service Discovery

Hello, Blazkowicz, you wrote: B> It is all it is already implemented in the client code? No, but it is necessary to implement all the same since application should inform on itself in service discovery. Other variants a-lja the centralized directory of services demands more time for service. Implementation is simple, at the consul is REST API. However, if at you not the cloud (all servers are reliable, applications  on one, instead of a cluster) and does not frighten directory support, is possible simply  XML with the directory of services on one of machines on HTTP and to draw out it from other applications. B> why? Need to twist DNS at clients only complicates . At us a shrink-wrapped software product. I am sorry, if at once did not specify. Especially, only to twist not at clients, and at itself. The circuit approximately the such: DNS specifies in 1-2 machines with nginx, and they  requests on necessary . A simple and reliable method to provide  and reconfiguration possibility . The interlayer in a type nginx is necessary, since changes in DNS are applied not at once + at equalization quite often more difficult behavior, than idle time DNS round-robin is necessary.

17

Re: Service Discovery

Hello, elmal, you wrote: E> At me too the box. Thus a cloud . Accordingly it is necessary on necessary  to deliver corresponding  and the docker to adjust it as it is necessary. Administrators for a week consulted With me. And further it is unrolled on time two and it is fine . As I will complete at last, there will be the full hair dryer . As a favor... You Can write small article about it here, for example? Would be  to esteem about something similar.

18

Re: Service Discovery

Hello, Blazkowicz, you wrote: AD>> It is possible to take a piece therefrom, namely SSDP (Simple Service Discovery Protocol). B> Yes, after  I at it am declined. But wanted to consult. Like as, microservices  . Someone can understood. I want to remind that UPnP, SSDP, etc. use . And if the network in your office is broken into subnets connected to routers,  from one subnet will get hardly to other subnet.

19

Re: Service Discovery

Hello, Kernan, you wrote: K> As a favor... You Can write small article about it here, for example? Would be  to esteem about something similar. Strongly laziness. Though also the pilot cluster project it was possible to launch with all it but while it is more experiment, at me even the necessary amount  is not present for a cluster. If I will ripen on article, it when will manage to be launched the project more seriously, and when all the same it will be possible to transplant all office on it. That there was a normal level of the competence at me, considering current level of bureaucracy on the coordination of tasks, some years still is required, though the pilot would be possible to launch normally. Actually and an Internet to line  on this business. In the core it is calculated for the ready torn clouds. In my case all differences - a cloud unrolled. I unrolled not, I by the machine at first delivered myself for experiments, then gave to administrators of the reference as well as what to put and told what machines to unite in a cluster. Then started to unroll any more on a local computer, and on a cluster - it is a little  for there virtual ip containers appeared in different networks, and hazelcast a horse-radish did not find the automatic machine, at me not distributed application, and n applications independent, fulfilling one and too turned out, and not in a state to load the data for they did not get in . But  found the decision, and through kubernetes api learned at start the list  on which application is torn, and these ip fed hazelcast explicitly - thus earned. At hazelcast by the way support - I am simple in delight, answer almost instantly is direct in a chat if what problem - speak where in the code to look a debugger, and developers answer. Like it was possible to try for the same mesos to use, but it will seemed that kubernetes is easier for development and there more from a box is. What that of the basic jambs at kubernetes did not find, though and on bugs . Not so to me truth clearly as update most kubernetes to do that all to devils did not fly, but this business of the future.